Research On Word Sense Disambiguation Based On DBN

There are a number of ambiguous words in Chinese vocabulary.Ambiguous words bring great convenience to the use of natural language.Ambiguous words also bring some difficulties to the understanding and translation of natural language.Word sense disambiguation is to determine the true meaning of ambiguous words according to the context of ambiguous words in the context.With the rise of artificial intelligence,word sense disambiguation has been applied more and more in many new and high-tech fields,which has become an important problem to be solved in natural language processing.Through the research on the domain knowledge of word sense disambiguation combined with the relevant knowledge of machine learning,this paper proposes a word sense disambiguation method based on Deep Belief Network(DBN),compares the traditional word sense disambiguation method based on Bayesian model,and compares the word sense disambiguation method based on RBM model.The method proposed in this paper has strong classification ability,and the disambiguation accuracy of ambiguity words is much higher than that of traditional methods.The research content of this paper is divided into the following parts:Firstly,this paper introduces the research purpose and significance of word sense disambiguation,analyzes the research status and development trend at home and abroad,and introduces some authoritative methods of word sense disambiguation at home and abroad as well as the main research content of the subject.Secondly,some basic knowledge of natural language processing is studied,and synonym word forest is introduced in detail.The selection process of disambiguation features of Bayesian classifier,RBM classifier and DBN classifier is described in detail.For Bayesian classifier,word form and part of speech in an ambiguous word's left and right four adjacent word units are used as disambiguation features to determine its semantic category.For RBM classifier and DBN classifier,word form,part of speech and semantic categories in an ambiguous word's left and right four adjacent word units are used as disambiguation features to determine its semantic category.Finally,DBN is used to construct word sense disambiguation model.Combining training corpus of Semeval-2007: Task#5 and semantic annotation corpus in Harbin Institute of Technology to optimize the DBN's parameters.Testing corpus of Semeval-2007: Task#5 was used to test the word sense disambiguation classifier.In this paper,the Bayesian classifier and RBM classifier as a reference,a total of three groups of experiments were conducted.Through experimental comparison,the disambiguation ability of DBN classifier is higher than that of Bayesian classifier and RBM classifier.